Research Article

Received: 23 October 2010, Revised: 11 January 2011, Accepted: 17 January 2011, Published online in Wiley Online Library: 17 February 2011

(wileyonlinelibrary.com) DOI: 10.1002/cem.1385

A quantum chemical and chemometric studyof sesquiterpene lactones with cytotoxicityagainst tumor cellsFrancisco F. P. Arantesa, Luiz C. A. Barbosaa*, Celia R. A. Malthaa,Antonio J. Demunera, Paulo H. Fidenciob and Jose Walkimar M. Carneiroc

The semi-empirical molecular orbital method PM6 w

J. Chemom

as employed to calculate a set of molecular descriptors of 20sesquiterpene lactones (SQLs) with cytotoxicity against HL-60 (leukemia) tumor cells. The principal componentanalysis (PCA) and hierarchical cluster analysis (HCA) methods were used to obtain possible relationships between thecalculated descriptors and the biological activity of the lactones. Four descriptors were identified as responsible forthe separation between the active and inactive compounds: EHOMO (highest occupied molecular orbital energy); Q11(net atomic charge on C11); Q12 (net atomic charge on C12) and Q13 (net atomic charge on C13). These resultsindicated that the presence of the a-methylene-g-lactone group has a significant role in themechanism by which SQLsexert their biological activities. Copyright � 2011 John Wiley & Sons, Ltd.

Keywords: sesquiterpene lactones; a-methylene-g-lactone; cytotoxicity; quantum chemical; chemometrics

* Correspondence to: L. C. A. Barbosa, Department of Chemistry, FederalUniversity of Vicosa, Av. P.H. Rolfs, S/N, CEP 36570-000, Vicosa, MG, Brazil.E-mail: lcab@ufv.br

a F. F. P. Arantes, L. C. A. Barbosa, C. R. A. Maltha, A. J. Demuner

Department of Chemistry, Federal University of Vicosa, Av. P.H. Rolfs, S/N, CEP

36570-000, Vicosa, MG, Brazil

b P. H. Fidencio

Department of Chemistry, Federal University of Jequitinhonha and Mucuri

Valleys, Campus JK, No 5000, Bairro Alto da Jacuba, CEP 39100-000,

Diamantina, MG, Brazil

c J. W. M. Carneiro

Department of Inorganic Chemistry, Federal University Fluminense, Outeiro

de Sao Joao Batista, s/n, Centro, CEP 24020-141, Niteroi, RJ, Brazil 4

1. INTRODUCTION

Sesquiterpene lactones (SQLs) are an important class of naturalproducts found in plants of the Asteraceae family, known for theirvarious biological activities such as anti-inflammatory, phytotoxic,antiprotozoal and cytotoxicity against different tumor cell lines[1–9]. In most cases, the biological activity of SQLs is related to thea-methylene-g-lactone functionality, which is prone to react withsuitable nucleophiles as sulfhydryl groups of cysteine in a Michaeladdition type reaction [10–16].Two different situations can be evaluated when a structure–

activity relationship (SAR) study is performed: the active site ofthe receptor is known or unknown. For the first case, informationabout the receptor site can be obtained from molecularmodeling, X-ray analysis or nuclear magnetic resonance (NMR)studies. When the active site is unknown, SAR or quantitativestructure–activity relationship (QSAR) techniques can be appliedto a series of similar compounds with known biological activitypreviously obtained [17–22].SAR studies have been proven to be helpful in the under-

standing of the influence of molecular properties on thebiological activity presented by several kinds of compounds.Quantum chemical parameters of molecules and even of theinteracting molecular systems can, in principle, express allelectronic properties related to the molecular interactions. Thus,SAR studies using quantum chemical parameters have becomeimportant in qualitative and quantitative analyses of three-dimensional molecular interactions [17–22].Continuing our efforts to prepare compounds with high

cytotoxic activity [16,18,23–27], we describe herein a study of therelationship between selected molecular parameters (descrip-tors) and cytotoxicity of a set of SQLs. The semi-empiricalPM6 method was employed to calculate atomic and molecular

etrics 2011; 25: 401–407 Copyright � 2011 J

descriptors of 20 SQLs reported in our previous works [16,28] ascytotoxic agents.The descriptors (variables) in this work were chosen taking

into account three classes of variables: electronic, steric andhydrophobic, as they represent the possible molecular inter-actions between the SQLs and the biological receptor. Theprincipal component analysis (PCA) and the hierarchicalcluster analysis (HCA) were employed to obtain a relationshipbetween the calculated variables and the cytotoxicity againstHL-60 (leukemia) tumor cells.

2. MATERIALS AND METHODS

2.1. Compounds

The chemical structures of the 20 SQLs studied in this work arepresented in Scheme 1. The numbering adopted for the carbon

ohn Wiley & Sons, Ltd.

01

F. F. P. Arantes et al.

402

atoms is shown in the structure of compounds 1, 2, 3 and 5.The respective IC50 values (concentration to exert 50% growthinhibition against HL-60 tumor cells) of all the SQLs studied areshown in Table I.

O

O

OO

O

O

OAc

O

O

O

O

O

O

SePh

O

O

O

OAc

O

O

OAc

HO

O

O

HOOC

O

O

OH

O

O

O

O

O

O

O

HOOC

O

(1) (2)

(5)

(6)

(9) (10)

(13))41(

(17) (18)

1113

12

1

2

3

45

12

3

45

123

45

Scheme 1. Structure of the 20 se

wileyonlinelibrary.com/journal/cem Copyright � 2011 John

2.2. Calculation of the atomic and molecular descriptors

The geometries of the 20 SQLs were fully optimized using themolecular mechanics force field (MMFF) method [29]. When

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

OH

O

O

O

O

OH

O

O

HOOC

O

O

O

O

O

OH

HO

O

O

O

SePh

+

(3)(4)

(7) (8)

(11) (12)

)61()51(

(19)

(20)

12

34

5

squiterpene lactones studied.

Wiley & Sons, Ltd. J. Chemometrics 2011; 25: 401–407

Table I. Values of the four most important properties that classify the 20 sesquiterpene lactones and their cytotoxicity (IC50) againstHL-60 (leukemia) tumor cells. EHOMO is the energy of the highest occupied molecular orbital. Q11, Q12 and Q13 are the atomiccharges on carbon atoms 11, 12 and 13, respectively

Compounds EHOMO (eV) Q11 Q12 Q13 IC50 (mM)

1 �10.008 �0.2033 0.6032 �0.4572 >1002 �10.062 �0.2070 0.6075 �0.4574 >1003 �9.229 �0.2119 0.6133 �0.4512 >1004 �10.037 �0.2086 0.6072 �0.4570 >1005 �10.013 �0.2022 0.6052 �0.4572 80.16

65.68–91.316 �8.921 �0.0970 0.6023 �0.4757 >1007 �10.045 �0.1537 0.6112 �0.2374 1.14

0.23–2.778 �10.041 �0.1528 0.6133 �0.2377 2.30

1.87–2.849 �10.089 �0.1574 0.6149 �0.2333 1.60

1.09–2.3510 �9.704 �0.2085 0.6094 �0.4565 >10011 �9.688 �0.2102 0.6092 �0.4562 >10012 �8.974 �0.0899 0.5953 �0.4752 8.73

6.98–10.9713 �9.255 �0.1570 0.6197 �0.2363 8.70

6.87–11.0714 �10.059 �0.1579 0.6144 �0.2346 5.70

4.56–6.8415 �9.683 �0.1509 0.6148 �0.2379 11.90

8.62–16.4116 �10.142 �0.2095 0.6049 �0.4453 >10017 �9.981 �0.2032 0.6044 �0.4451 >10018 �9.391 �0.2009 0.6067 �0.4570 >10019 �10.155 �0.2148 0.6087 �0.4455 >10020 �9.410 �0.1528 0.6144 �0.2343 1.45

1.45–1.81

Cytotoxicity of new sesquiterpene lactones

necessary, several conformations were calculated for a givencompound and only the most stable one was consideredfurther. The molecular descriptors were calculated using thesemi-empirical PM6 method [30], based on the most stableconformation of each derivative. MMFF and PM6 calculationswere done using the PC Spartan Pro [31] and MOPAC 2009 [32]software, respectively.The following descriptors were calculated:

(1) E

J. C

lectronic descriptors

� The energy of the highest occupied molecular orbital(HOMO energy) and of the lowest unoccupied molecularorbital (LUMO energy);

� Mulliken electronegativity (X), obtained according to thefollowing equation: X¼ (EHOMOþ ELUMO)/2;

� Electron affinity (EA), obtained as (�ELUMO);� Dipole moment (m);� Heat of formation (DHf );� Total energy (Et);� Electronic energy (Eel);� Net atomic charge on the carbon atoms C1 (Q1), C2 (Q2), C3(Q3), C4 (Q4), C5 (Q5), C11 (Q11), C12 (Q12) and C13 (Q13).

hem

4

(2) S teric descriptors

ometrics 2011; 25: 401–407 Copyright � 2011 John Wiley &

� Molecular area (MA);� Molecular volume (MV).

So

(3) H

ydrophobic descriptor

� Partition coefficient (log P).� The partition coefficients were calculated using the PCSpartan Pro software. The statistical analysis (PCA and

HCA) was performed using the MATLAB 6.0 program [33].

3. RESULTS AND DISCUSSION

3.1. PCA

PCA is frequently employed to reduce the dimensionality ofa multidimensional system. The main objective of PCA is tocompress data into a small group of new variables, which arelinear combination of the original variables that maximizedescription of the total variance data. Geometrically, thistransformation represents rotation of the original coordinatesystem, so that the direction of the maximum residual variance isgiven by the first principal component axis. The second principalcomponent axis, orthogonal to the first one, has the secondmaximum variance and so on [34,35].

ns, Ltd. wileyonlinelibrary.com/journal/cem

03

Table II. Loading vectors for the first four principal com-ponents

Variable PC1 PC2 PC3 PC4

EHOMO �0.1145 0.1904 0.6609 0.7169Q11 0.6807 0.7080 �0.1710 0.0783Q12 0.6642 �0.4881 0.5138 �0.2379Q13 0.2871 �0.4735 �0.5197 0.6507

F. F. P. Arantes et al.

404

The PCA method was used in order to obtain a separationof the set of compounds in two groups (actives and inactives),according to the calculated molecular descriptors. In addition,and most importantly, following the classification PCA is ableto indicate the relevance of the selected molecular descriptorto differentiate between the active and inactive compounds.Therefore, these results give indication of the molecular des-criptors responsible for the activities. Before applying the PCAmethod, each variable was auto-scaled so that they could becompared to each other on the same scale. After scaling, severalattempts were made to obtain a good classification of the setof compounds. The best separation was obtained with fourvariables (see Table I) out of the 19 we initially had: EHOMO (highestoccupied molecular orbital energy); Q11 (net atomic charge onC11); Q12 (net atomic charge on C12) and Q13 (net atomic chargeon C13). This suggests that the other variables, including thecharges on the other sp2 carbon atoms, are not important forclassifying these compounds according to their cytotoxicities.Further analysis was done using only this subset of moleculardescriptors.Using this reduced set of variables, the total variance of the

original data set is represented by four PCs as follows: PC1¼46.09%, PC2¼ 39.89%, PC3¼ 13.26% and PC4¼ 0.76%. Anumber of score plots were examined and the most informativeone (PC1� PC2) is presented in Figure 1. This projection keeps85.98% of the total variance of the original data set and can beexpected to provide a reasonably accurate representation of thehigher order space.Figure 1 shows that the SQLs analyzed are separated into two

groups, A and B. Group A contains the SQLs 1–6, 10–12, 16–19,with lower degree of cytotoxicity (IC50> 80.16mM) against HL-60tumor cells, except compound 12 with IC50 of 8.73mM (Table I).Group B consists of the SQLs (compounds 7, 8, 9, 13, 14, 15, 20)with higher cytotoxicity (1.14< IC50< 11.90mM) against HL-60

-2.5 -2 -1.5 -1 -0.5-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

1

2

3

4 5

6

10 11

12

16

17

18

19

PC 1 (

PC

2 (

39.8

9%)

Lower Activity (Group A)

Figure 1. Plot of the score vectors of the principal components (PC1� PC2)

The PCA separates the compounds into two groups: lower activity (group A

wileyonlinelibrary.com/journal/cem Copyright � 2011 John

tumor cells (Table I). Additionally, it can be seen that GroupsA and B are separated mainly along PC1. The loading vectors forthe selected variables in PC1, PC2, PC3 and PC4 are given inTable II. Figure 2 displays the plot of the loading vectors for thefirst two PCs (PC1� PC2). According to Figure 2, the EHOMO

descriptor is responsible for describing the inactive compounds(PC1< 0) and the Q11, Q12 and Q13 descriptors are responsiblefor describing the active ones (PC1> 0).According to Table II, PC1 can be expressed through the

following equation:

PC1 ¼ 0:6807 Q11½ � þ 0:6642 Q12½ �

þ 0:2871 Q13½ ��0:1145 EHOMO½ �

From this equation, it can be seen that for a given SQL tobecome active (PC1> 0) it must have more negative values forEHOMO (note that EHOMO has negative values), less negative valuesof Q11 and Q13, together with more positive values of Q12.The energy of the frontier orbitals is an important property

in several chemical and pharmacological processes, and thereason for this is the fact that this property gives informationon the electron-donating and electron-accepting character of a

0 0.5 1 1.5 2 2.5

7

8 9

13

14

15

20

46.09%)

Higher Activity (Group B)

for the 20 sesquiterpene lactones with cytotoxicities against tumor cells.

) and higher activity (group B).

Wiley & Sons, Ltd. J. Chemometrics 2011; 25: 401–407

Figure 2. Plot of the first two loadings vectors (PC1 and PC2) of the variables responsible for the separation of the active and inactive compounds.

Cytotoxicity of new sesquiterpene lactones

compound, i.e. on the formation of a charge transfer complex.The above equation indicates that the energy of HOMO (EHOMO)for the active compounds must present lower (more negative)values than the inactive compounds. This means that theactive compounds are not good electron donors when comparedto the inactive ones, which is consistent with the fact that, in mostcases, the biological activity of SQLs is related to reaction ofthe a-methylene-g-lactone group with suitable nucleophiles,e.g. sulfhydryl groups of cysteine, in a Michael addition typereaction [36]. This is also a consequence of the higher positive

0 0.5 1 1.50

2

4

6

8

10

12

14

16

18

20

Simila

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

Lactoneof

number

Figure 3. Dendrogram obtained with hierarchical cluster analysis for the 20

classifies the compounds into two groups: lower activity (group A) and high

J. Chemometrics 2011; 25: 401–407 Copyright � 2011 John Wil

charge on C12. Thus, bioactive SQLs are, generally, good electronacceptors.

3.2. HCA

In the HCA methodology, distances between pairs of samples arecalculated and compared. Small distances between samplesimply that they are similar. On the other hand, dissimilar sampleswill be separated by relatively large distances. HCA starts with

2 2.5 3 3.5

rity

Higher activity

Lower activity (Group A)

(Group B)

sesquiterpene lactones with cytotoxicities against tumor cells. The HCA

er activity (group B).

ey & Sons, Ltd. wileyonlinelibrary.com/journal/cem

405

F. F. P. Arantes et al.

406

each sample defined as its own cluster, then similar samples aregrouped together to form new clusters until all samples arepart of a single cluster. The main purpose of HCA is to representdata in a manner that emphasizes natural groupings assigning,thus, categories to which samples belong. The visualization ofthe groups corresponding to different classes is achieved in theform of dendrograms where the vertical lines represent thecompounds and the horizontal lines represent the similaritybetween them [20].The results obtained with the HCA analysis are displayed in the

dendrogram shown in Figure 3. The same descriptors selectedby PCA were used (EHOMO, Q11, Q12, Q13). Figure 3 shows thatthe 20 SQLs are separated in the two groups A and B, exactly asobserved in the PCA analysis. Thus, the EHOMO, Q11, Q12 andQ13 descriptors are confirmed as the most important ones forclassification of the SQLs in inactive or active compounds againstHL-60 tumor cells.

4. CONCLUSIONS

PCA and HCA showed that the 20 SQLs studied can be classifiedinto two groups: active (group A) and inactive (group B) againstHL-60 tumor cells. The electronic descriptors EHOMO, Q11, Q12 andQ13 are the most important for the separation between activeand inactive molecules. This indicates that electronic effectsplay an important role in the understanding of cytotoxicity of theSQLs against tumor cells, while steric (MA and MV) andhydrophobic (LogP) descriptors are not important for classifyingthese compounds according to their cytotoxicities. The PCAanalysis indicates that compounds with more negative EHOMO,less negative Q11 and Q13 and more positive values of Q12 arethe most active ones. These results reinforce the fact that thepresence of the a-methylene-g-lactone group, prone to reactwith nucleophiles in a Michael addition reaction, has a significantrole in the mechanism by which SQLs exert their biologicalactivities. On the basis of the results of this study, new SQLs canbe designed which will probably show higher cytotoxicitiesagainst tumor cells.

Acknowledgements

The authors are grateful to the following Brazilian agencies:Conselho Nacional de Desenvolvimento Cientıfico e Tecnologico(CNPq) for research fellowships (AJD, CRAM, LCAB and JWMC),Fundacao de Amparo a Pesquisa de Minas Gerais (FAPEMIG),Coordenacao de Aperfeicoamento de Pessoal de Nıvel Superior(CAPES, PROCAD program, grant 23038.022059/2008-92) andFINEP for financial support. The authors also thank Prof. ClaudiaO Pessoa for collaboration on this project.

REFERENCES

1. Hehner SP, Heinrich M, Bork PM, Vogt M, Ratter F, Lehmann V,Schulze-Osthoff K, Droge W, Schmitz ML. Sesquiterpene lactonesspecifically inhibit activation of NF-kB by Preventing the Degradationof IkB-a and IkB-b. J. Biol. Chem. 1998; 273: 1288–1297.

2. Schmidt TJ, Brun R, Willuhn G, Khalid SA. Anti-trypanosomal activity ofHelenalin and some structurally related sesquiterpene lactones.Planta Med. 2002; 68: 750–751.

3. Zhang S, Won YK, Ong CN, Shen HM. Anticancer potential of sesqui-terpene lactones: bioactivity and molecular mechanisms. Curr. Med.Chem. Anticancer Agents 2005; 5: 239–249.

wileyonlinelibrary.com/journal/cem Copyright � 2011 John

4. Barbosa LCA, Demuner AJ, Borges EEL, Mann J. Synthesis and evalu-ation of the plant growth regulatory activity of 8-oxabicyclo[3.2.1]oct-6-en-3-one derivates. J. Braz. Chem. Soc. 1997; 8: 19–27.

5. Barbosa LCA, Maltha CRA, Borges EEL. Sıntese e avaliacao da atividadefitotoxica de lactonas derivadas do 2,4-dimetil-8-oxabiciclo[3.2.1]oct-6-en-3-ona. Quım. Nova 2002; 25: 203–208.

6. Macıas FA, Galindo JCG, Castellano D, Velasco RF. Sesquiterpenelactones with potential use as natural herbicide models. 2. Guaiano-lides. J. Agric. Food Chem. 2000; 48: 5288–5296.

7. Macıas FA, Fernandez A, Varela RM, Molinillo JMG, Torres A, AlvesPLCA. Sesquiterpene lactones as allelochemicals. J. Nat. Prod. 2006;69: 795–800.

8. Barbosa LCA, Costa AV, Pilo-Veloso D, Lopes JLC, Hernandez-TerronesMG, King-Diaz B, Lotina-Hennsen B. Phytogrowth-inhibitory lactonesderivatives of glaucolide B. Z. Naturforsch. 2004; 59: 803–805.

9. Alvarenga ES, Barbosa LCA, Saliba WA, Arantes FFP, Demuner AJ.Sıntese e avaliacao da atividade fitotoxica de derivados daa-Santonina. Quım. Nova 2009; 32: 401–406.

10. Polo LM, Castro CM, Cruzado MC, Collino CJG, Cuello-Carrion FD,Ciocca DR, Giordano OS, Ferrari M, Lopez LA. 11,13-Dihydro-dehydrol-eucodine, a derivative of dehydroleucodine with an inactivatedalkylating function conserves the anti-proliferative activity inG2 but does not cause cytotoxicity. Eur. J. Pharmacol. 2007; 556:19–26.

11. Kupchan SM, Eakin MA, Thomas AM. Tumor inhibitors. 69. Structure-cytotoxicity relations among the sesquiterpene lactones. J. Med.Chem. 1971; 14: 1147–1152.

12. Lee KH, Wu YS, Hall IH. Antitumor agents. 25. Synthesis and antitumoractivity of uracil and thymine alpha-methylene-gamma-lactones andrelated derivatives. J. Med. Chem. 1977; 20: 911–914.

13. Nakagawa Y, IinumaM, Matsuura N, Yi K, Naoi M, Nakayama T, NozawaY, Akao Y. A potent apoptosis-inducing activity of a sesquiterpenelactone, arucanolide, in HL60 Cells: a crucial role of apoptosi-s-inducing factor. J. Pharmacol. Sci. 2005; 97: 242–252.

14. Rozalski M, Krajewska U, Panczyk M, Mirowski M, Rozalska B, Wasek T,Janecki T. Synthesis and biological evaluation of 4-methyl-ideneisoxazolidin-5-ones—A new class of highly cytotoxic a-methyl-idene-g-lactones. Eur. J. Med. Chem. 2007; 42: 248–255.

15. Macıas FA, Galindo JCG, Massanet GM. Potential allelopathic activity ofseveral sesquiterpene lactone models. Phytochemistry 1992; 31:1969–1977.

16. Arantes FFP, Barbosa LCA, Alvarenga ES, Demuner AJ, Bezerra DP,Ferreira JRO, Costa-Lotufo LV, Pessoa C, Moraes MO. Synthesis andcytotoxic activity of a-santonin derivatives. Eur. J. Med. Chem. 2009;44: 3739–3745.

17. Costa MCA, Barata LES, Takahata Y. SAR analysis of synthetic neo-lignans and related compounds which are anti-leishmaniasis activecompounds using pattern recognition methods. J. Mol. Struct. (Theo-chem.) 1995; 340: 185–192.

18. Alvarenga ES, Silva SA, Barbosa LCA, Demuner AJ, Parreira AG, RibeiroRIMA, Marcussi A, Ferreirea JMS, Resende RR, Granjeiro PA, Silva JA,Soares AM, Marangoni S, Silva SLD. Synthesis and evaluation ofsesquiterpene lactone inhibitors of phospholipase A2 from Bothropsjararacussu. Toxicon 2011; 57: 100–108.

19. Camargo AJ, Mercadante R, Honorio KM, Alves CN, Da Silva ABF. Astructure–activity relationship (SAR) study of synthetic neolignansand related compounds with biological activity against Escherichiacoli. J. Mol. Struct. (Theochem.) 2002; 583: 105–116.

20. Souza J, Santos RHA, Ferreira MMC, Molfetta FA, Camargo AJ, HonorioKM, Da Silva ABF. A quantum chemical and statistical study offlavonoid compounds (flavones) with anti-HIV activity. Eur. J. Med.Chem. 2003; 38: 929–938.

21. Yang H, Chen G, Li Y. A quantum chemical and statistical study ofganoderic acids with cytotoxicity against tumor cell. Eur. J. Med. Chem.2005; 40: 972–976.

22. Molfetta FA, Bruni AT, Honorio KM, Da Silva ABF. A structure–activityrelationship study of quinone compounds with trypanocidal activity.Eur. J. Med. Chem. 2005; 40: 329–338.

23. Teixeira RR, Barbosa LCA, Maltha CRA, Rocha ME, Bezerra DP, Costa-Lotufo LV, Pessoa C, Moraes MO. Synthesis and cytotoxic activity ofsome 3-benzyl-5-arylidenefuran-2(5H)-ones. Molecules 2007; 12:1101–1116.

24. Pessoa C, Silveira ER, Lemos TL, Wetmore LA, Moraes MO, Leyva A.Antiproliferative effects of compounds derived from plants of North-east Brazil. Phytother. Res. 2000; 14: 187–191.

Wiley & Sons, Ltd. J. Chemometrics 2011; 25: 401–407

Cytotoxicity of new sesquiterpene lactones

25. Costa-Lotufo LV, Silveira ER, Barros MC, Lima MA, De MoraesME, De Moraes MO, Pessoa C. Antiproliferative effects of AbietaneDiterpenes from Aegiphila lhotzkyana. Planta Med. 2004; 70: 180–182.

26. Bezerra DP, Pessoa C, Moraes MO, Alencar NM, Mesquita RO, LimaMW, Alves AP, Pessoa OD, Chaves JH, Silveira ER, Costa-Lotufo LV. Invivo growth inhibition of sarcoma 180 by piperlonguminine, analkaloid amide from the Piper species. J. Appl. Toxicol. 2008; 28:599–607.

27. Kiran YB, Reddy CD, Gunasekar D, Reddy CS, Leon A, Barbosa LCA.Synthesis and anticancer activity of new class of bisphosphonates/phosphanamidates. Eur. J. Med. Chem. 2008; 43: 885–892.

28. Arantes FFP, Barbosa LCA, Maltha CRA, Demuner AJ, Costa PM, FerreiraJRO, Costa-Lotufo LV, Moraes MO, Pessoa C. Synthesis of novela-santonin derivatives as potential cytotoxic agents. Eur. J. Med. Chem.2010; 45: 6045–6051.

29. Halgren TA. Merck molecular force field. I. Basis, form, scope, para-meterization, and performance of MMFF94. J. Comput. Chem. 1996;17: 490–519.

J. Chemometrics 2011; 25: 401–407 Copyright � 2011 John Wil

30. Stewart JJP. Optimization of parameters for semiempirical methods V:modification of NDDO approximations and application to 70elements. J. Mol. Model. 2007; 13: 1173–1212.

31. Spartan PC Pro. Wavefunction, Inc., Irvine, CA, USA.32. Stewart JJP. MOPAC2009, Stewart Computational Chemistry, version

9.327W.33. Wise M, Gallagher NB. PLS Toolbox Ver. 2.0.1c, for use with MATLAB

6.0, 1999.34. Boon G, Langenaeker W, De Proft F, Hans DeWinter JP, Tollenaere P.

Systematic study of the quality of various quantum similarity descrip-tors. Use of the autocorrelation function and principal componentanalysis. J. Phys. Chem. A 2001; 105: 8805–8814.

35. Pinheiro JC, Ferreira MMC, Romero OAS. Antimalarial activity ofdihydroartemisinin derivatives against P. falciparum resistant tomefloquine: a quantum chemical andmultivariate study. J. Mol. Struct.(Theochem) 2001; 572: 35–44.

36. Mattos MC, Marzorati L. Aspectos Mecanısticos da Adicao de Michael.Quım. Nova 1999; 22: 710–714.

ey & Sons, Ltd. wileyonlinelibrary.com/journal/cem

407